File size: 501 Bytes
4245765 |
1 2 3 4 5 6 7 8 9 10 11 12 |
## RusEnQA
QA for Russian and Englisha based on the [rugpt3xl](https://huggingface.co/sberbank-ai/rugpt3xl) model
### About ruGPT-3 XL model
Model was trained with 512 sequence length using Deepspeed and Megatron code by SberDevices team, on 80B tokens dataset for 4 epochs. After that model was finetuned 1 epoch with sequence length 2048.
Note! Model has sparse attention blocks.
Total training time was around 10 days on 256 GPUs.
Final perplexity on test set is 12.05. Model parameters: 1.3B. |